Interview with Synthefy

The winning teams of the T Challenge 2024 have just been announced. First place went to Synthefy (USA), a spinout from Stanford University and the Unversity of Texas, Austin, with generative AI for the analysis of timeseries and network log data. Synthefy's platform incorporates multi-modal metadata and enables analyzing of timeseries by simple text prompts. Today we talked to Somi Agarwal, CEO of Synthefy.

How does it feel to win the T-Challenge award?

Somi Agarwal, CEO of Synthefy: It's an incredible feeling, and I’m still processing it. For me, those couple of days were really special for various of reasons. It’s my first visit to Germany, and receiving the award from the hands of Deutsche Telekom - which was our first customer when we started just nine months ago - serves as a confirmation of what we do. Winning the T Challenge shows that is it possible to add real value to a telco business at a global scale, even if you are a small company based in Austin, Texas, with a team of five. Despite our size, we’re driven and energetic, eager to make a significant impact.

How do you think winning the T-Challenge award will scale your business?

Somi Agarwal: I believe it will be transformative. We've talked to other telecom companies, and mentioning our work with DT immediately validates our capabilities. Solving network-resilience problems for the biggest telecom operator in the world demonstrates our potential to others. With 250 million customers globally, DT is a significant benchmark and a brand that brings more trust to Synthefy as a startup.

Along with the award, you received 150,000 euros. How do you plan to invest it?

Somi Agarwal: It’s an open question since we're still considering our options, but as a small company, this funding is crucial. It will help us grow, hire more people, and scale our AI data-driven solutions. Our goal is to integrate our product inside of the DT network, making it more robust and stable for your end customers.

Can you explain the core of your business at Synthefy? What drives you to innovate?

Somi Agarwal: We’re a team of young engineers and researchers passionate about technology. We develop advanced algorithms, neural networks, and deep learning models. Seeing our innovations solve real-world challenges is incredibly rewarding - that’s what keeps us going. We work on a couple of domains. We forecast sales and inventory while accounting for ads, website traffic, past sales, and macroeconomic conditions for e-commerce. We forecast energy demand patterns while accounting for weather, location, home type, and EV charging schedules for the energy sector and synthesize privacy-preserving medical ECG patterns based on patient demographics and health records for medical companies. But telecommunications focused on data is in our DNA.

Why did you choose to focus on that domain?

Somi Agarwal: Because data is the truth of the world. It reflects what has happened, and everything we can achieve starts with data. By analyzing it, we can make predictions, such as identifying potential failures. Understanding and leveraging relevant metadata is fundamental to forecasting and everything we do.

How can your work impact the telecommunications network and help Deutsche Telekom, also regarding AI as a main topic of T Challenge 2024?

Somi Agarwal: The impact of AI is enormous and transformative, especially in telecom. Our solutions aim to make telecom operations more efficient and robust. We synthesize a digital twin for a network based on application traffic types, network topology, cell location, or even system logs. Because of that, with the help of AI models, we can predict network failures - sometimes up to 90% accuracy even 10 hours before the incident happens. How does it work? We look at past data to learn how it failed before. Our AI learns from this to predict future failures, similar to how a child learns to walk by trying, failing, and learning from mistakes. Working with Deutsche Telekom we focused on predicting server failures that disrupt connections. We also work on identifying issues in base stations and other components of the network.

What models have you developed for these predictions?

Somi Agarwal: Our key novelty is the algorithms that we’ve created through our PhD research. It’s the world's first foundation model for multi-modal time series data. These in-house models are tailored for telecom and time series data, providing unique solutions not available elsewhere.

How will you measure your success in the coming years?

Somi Agarwal: Participating in this competition and meeting industry leaders has broadened our understanding of telecom challenges. We've learned about new problems we weren't aware of before. This experience will shape our strategy for the next few years, helping us focus on the most critical issues for DT and the telecom industry. Our measure of cusses will be a functional solution, that helps to reduce operational costs for your company, serving as a “Minority Report” for server infrastructure. Preventing something wrong before it happens.

Since you are located in Texas, will you also work with telecom companies in the US?

Somi Agarwal: Yes, we're in talks with some US companies, but DT has been the most forward-thinking partner we've encountered. You took a chance on us, and we've delivered results. We look forward to continuing our work with DT and applying our solutions to other telecoms.

We’re excited to see your future successes after the T Challenge award and wish you the best of luck.

Somi Agarwal: Thank you, it was a unique experience for us, and now it’s time to take Synthefy to another level.

More News

T Challenge 2024: Winner Announcement

The telecommunications landscape is rapidly evolving, and AI is at the forefront of this transformation. To accelerate this progress, Deutsche Telekom and T-Mobile US have ...
Read More

T Challenge 2024: Grand Jury Announcement

On June 10-11, 2024 Deutsche Telekom headquarters in Bonn, Germany, will be the stage for twelve finalist teams. They will present their innovative solutions on ...
Read More

T Challenge 2024: Pioneering AI Ideas for Telecommunications

Deutsche Telekom and T-Mobile US proudly announce the final twelve nominees for the T Challenge 2024 on AI for Telecommunications. The competition has seen an ...
Read More